W-band prototype of platelet feed-horn array for CMB polarisation measurements

Preprint typeset in JINST style - HYPER VERSION

o

(N

*sO

J2

O

W-band prototype of platelet feed-horn array for
CMB polarisation measurements

F. Del Torto a * M. Bersanelli , F. Cavaliere", A. De Rosa 6 , 0. D'Arcangelo',

C. Franceschet , M. Gervasi^, A. Mennella ", E. Pagana e , A. Simonetto c , A. Tartar^,

F. Villa 6 , M. ZannonK

"Dipartimento di Fisica, Universita degli Studi di Milano,
via Celoria 16, 20133 Milano, Italy,
b INAF-IASF Bologna,
Via Gobetti 101, 40129 Bologna, Italy,
X ; C IFP-CNR,

Vj ■ Via Cozzi 53, Milano, Italy

Dipartimento di Fisica, Universita degli Studi di Milano Bicocca,
Piazza delta Scienza 3, 20126 Milano, Italy,
"Independent consultant

f~^. ' E-mail: francesco.deltorto@mi.infn.it

in '

Abstract: We present the design and performance of a 2x2 prototype array of corrugated feed-
horns in W-band. The module is fabricated using a so-called "platelet" technique by milling Alu-
minum plates. This technique is suitable for low-cost and scalable high performance applications.
Room temperature Return Loss measurements show a low (<-30 dB) reflection over a 30% band-
width with a maximum matching of -42 dB at 100 GHz for all four antennas. Beam pattern mea-
^ surements indicate good repeatability and a low (-25 dB) sidelobe and crosspolarisation levels. This

work is particularly relevant for future Cosmic Microwave Background polarisation measurements,
which require large microwave cryogenic detector arrays coupled to high performance corrugated
feed horns.

KEYWORDS: Instruments for CMB observations, Microwave antennas, Detector design and
construction technologies and material, Polarisation.

* Corresponding author.

Contents

|l[ Introduction

g, The Prototype

[2.1| Electromagnetic design

2.2 Mechanical design

2.3 Materials

2.4 Manufacturing

g Testing

[3.1| Beam Patterns

3.2 Return Loss

3.3| Insertion Loss |12

f|. Conclusions [fj|

|5[ Acknowledgments

1. Introduction

Forty-five years after its discovery, the Cosmic Microwave Background (CMB) radiation is today
a very powerful tool of cosmological investigation.

The CMB originates from the decoupling of matter and radiation, occurred when the temper-
ature of the expanding universe fell below 3000 K. At this temperature the plasma could combine
into neutral hydrogen, thus reducing the Thomson scattering cross section and letting the radiation
propagate freely. Because of cosmic expansion the CMB can be observed today as a blackbody
radiation at the temperature T=2.725 ± 0.002 K ^ peaked in the microwave region of the electro-
magnetic spectrum.

The CMB temperature is anisotropic at the level of AT /T fa 10~ 5 on scales from 180 degrees
to about 5'. The angular temperature anisotropics depend on the initial conditions of the Universe
and on the physical processes that occurred from the Big Bang to the matter-radiation decoupling
and later via Sunyaev-Zel'dovich effect. Accurate measurements of these anisotropics, therefore,
provide a wealth of information on the universe history and composition (baryons, dark matter,
dark energy). CMB temperature anisotropics have been measured by many ground and balloon
experiments as well as by the NASA satellites COBE 1 and WMAP 2 (see [0] for an extensive review

http://aether.lbl. gov/ www/pro jects/cobe/
http : //map .gsfc.nasa. gov/

until WMAP). The third generation ESA satellite Planck 3 , launched in May 2009, will provide in
few years foreground-limited CMB temperature anisotropy measurements and open the path to
accurate measurements of CMB polarisation anisotropics, the future of CMB experiments.

Polarisation anisotropies, generated by Thomson scattering, can be represented on the celestial
sphere by two modes, named E-mode and B-mode, according to their symmetry properties. The
detection of B-modes would provide an indirect signature of cosmological gravitational waves, that
are predicted by the Hot Big Bang model [|3|].

Many experiments are currently planned to accurately measure the CMB polarisation. First de-
tection of E-modes and its correlation with the temperature anisotropy was published in 2003 by the
DASI collaboration H], while the most up-to-date measurements come from WMAP [|5j|. Because
only 10% of the CMB is polarised, polarisation anisotropies are smaller by a factor 10 compared
to temperature anisotropies. E-modes, in particular, are at the /xK level, while for B-modes the
signal is expected to be much weaker and not even well theoretically constrained, although it is not
exptected to be larger than some fraction of /iK ^.

Detection of such small signals call for ultra-high sensitivity instruments constituted of large
detector arrays of hundreds and even thousands of channels. A key element is represented by
the availability of high performance antennas which represent the interface between the sky (or
telescope) and the receiver. Precision polarisation measurements require optical components with
highly symmetrical optical response and losses not greater than ~ 1%. Corrugated feed-horns match
these requirements and are commonly used in most CMB experiments [cfr. WMAP and Planck].
Typical manufacturing techniques include direct machining, up ~40 GHz, and electroforming, at
higher frequencies. These methods can be limited by time and cost when applied to the produc-
tion of large detector arrays of hundreds of feed-horns, as requested by future CMB polarisation
experiments, so cheaper and quicker techniques are required.

In this work we investigate the validity of a technique, named "platelet" [fTp, [Q], [[0]], for
the mass production of corrugated feed-horn arrays. In particular we focus on three key factors:
(i) mechanical feasibility with reduced costs and time, (ii) reliability in standard and cryogenic
conditions and (iii) repeatability of the electromagnetic horn performance. Recently the QUIET 4
experiment started CMB polarisation measurememnts using, for the first time, feed-horn array
realized with the platelet technique |§j.

The platelet technique consists in the overlap of properly machined metal sheets so that the
feed-horn is then realized in layers. This method is relatively cheap, with a cost saving of about 90%
with respect to the electroforming and allows quick manufacturing of feed-horn arrays by drilling
several holes in each plate. A complete corrugation, with a tooth and a groove, can be obtained
in each plate. A complex focal surface can be obtained with a series of modules properly tilted to
reproduce the required geometry. Particular care is required in the plate assembly alignment system
and for mass/volume control in order to minimize the weight.

The advantages of this technique go beyond its application to CMB measurements and can be
exploited in a wide range of mm-wave and sub-mm astronomical instruments. In particular, platelet
techniques may prove appropriate for the new generation of large interferometric arrays, such as

http://planck.esa.int
http : //quiet . uchicago . edu

2-

ALMA, which require high performance antennas (including polarisation properties) over a large
number of channels.

2. The Prototype

To validate the "platelet" technique we have realized a prototype of 2x2 array of corrugated feed-
horn in W band (75-1 10 GHz), named WCAM01.

2.1 Electromagnetic design

The antenna electromagnetic design started form an already existing one []9|]. The design was
scaled to W band and the corrugation step was fixed to 1 mm in order to use standard aluminum
plates. The circular waveguide radius at the throat region was adapted by means of three steps
from 1.87 mm to 1.49 mm in order to match the circular-rectangular transition. In addition a 4 mm
long straight section has been added to guarantee the propagation of the TEH mode at the throat.
Because the main objective of this development was to test the manufacturing technique, we did
not perform any further optimizations of the electromagnetic design.

The horn aperture is D = 11.26 mm, giving a FWHM (~ A/D) of ~ 18 degrees at 100 GHz.
Its length is 41 mm with a flare angle of 12.6 degrees. Each corrugation tooth and groove is
0.3 mm and 0.7 mm wide, respectively. The corrugation depth is ~ A/4 at the aperture and linearly
increases to ~ A/2 in the last seven corrugations. A section representing the horn profile is shown
in Figure |].

±

-b

0,79

<-

Figure 1. Section of a corrugated feed-horn of the WCAM01 prototype.

The beam pattern simulations were obtained with the SRSR-D© France Telecom software.
The software is designed to simulate the electromagnetic performance of structures with cylin-
drical symmetry and consisting of perfect electrical conductor parts and homogeneous dielectric

3-

domains, therefore neglecting the metal resistive losses (i.e. the insertion loss). These restrictions
result in some limitations but allow very fast and reliable simulations. The radiation patterns have
been calculated at 90 GHz, 92.5 GHz, from 95 GHz to 105 GHz with a step of 0.25 GHz, then at
107.5 GHz and 110 GHz. Return loss was simulated with CST Microwave Studio© software.

2.2 Mechanical design

The mechanical design was conceived with the driving idea to obtain scalable modules suitable to
realise large feed-horn arrays. The prototype was manufactured (Figure ^) with a CNC machine
tool at the Physics Department workshop of the University of Milan.

The array has a square section of 70x70 mm, it is 44 mm high and composed by:

• one 7-mm thick base plate,

• thirty-three 1 -mm thick plates,

• one 4-mm thick head plate.

The plates are aligned by means of 8 rectified pins, alternated in pairs, and assembled together
with 9 rods. In this prototype the rod number and displacements have not been optimized for high
packing efficiency (horn aperture area/total area = 0.08) and lightweight. This will be the objective
of future studies aiming at designing more compact and light modules. The base plate is 7 mm thick
in order to host the three following elements: (i) the three steps required to reach the optimal exit
diameter of 2.98 mm, (ii) the anchorage holes (on the external side) for screws and pins for the
standard UG387 flanges, and (iii) the holes for the first alignment pins, that must be deep enough
to ensure that they are perpendicular to the plate. The 4 mm thickness of the head plate is required
to hide the rod nuts.

The feed-horn positioning and the interfaces were not designed for a particular type of detector
but just compatible with standard WR-10 waveguide flanges. In principle the array could feed both
bolometric detectors (with an appropriate thermal brake) and MMIC amplifier based detectors.

2.3 Materials

The chosen materials have been two aluminum alloys: A16082 Al-Mg-Si-Mn (Anticorodal) for the
plates, and A170750 Al-Zn-Mg-Cu (Ergal) for the rods. The pins and the nuts are made of stainless
steel.

The main characteristics of these aluminum alloys are their low specific weight (2.69 g/cm 3
for the Al 6082, 2.81 g/cm 3 for the Al 7075), their excellent workability and corrosion resistance,
their ability to be anodized and their good strength. The main difference between A16082 and
A17075 is the tensile strength that is respectively 310 N/mm 2 and 560 N/mm 2 . Moreover their
thermal expansion coefficients are quite similar: 23.4 for the A16082 and 23.5 for the A17075, thus
allowing to apply a great rod traction to properly bind the plates, and to stabilize the prototype with
respect to thermal contractions.

2.4 Manufacturing

Plates 1 through 33 were cut from 1 mm thick aluminum sheets and worked with the CNC machine
tool "Gualdoni GV-94". The base-plate and head-plate were milled from a 10 mm thick aluminum
slab and worked with the same machine. The rods were obtained by lathing 8 mm diameter Ergal
bars, threaded at their ends.

The corrugations were obtained by milling circular hollows concentric to the holes (Figure
|2|), with the difference in diameter corresponding to the corrugation depth, and the cave depth
(0.7 mm) corresponding to the the corrugation groove thickness. This process requested partic-
ular care to ensure the orthogonality between the plates and the mechanical tool. This condition
was guaranteed by a custom-designed plate support. The maximum error obtained on the teeth
corrugations thickness was of ±30 jUm.

Finally, the plates were manually assembled and aligned with the steel pins, and tightened
with the Ergal bolts. The assembly by means of rods is very quick, simple and does not require
any facilities and/or machinery. Moreover the array can be disassembled and modified in order to
possibly change its properties or replace inaccurate and/or damaged parts. Compared to bonding,
our technique has the drawback that a smaller number of horns can be packed in a single module.
On the other hand its simplicity, the flexibility that comes from the possibility to disassemble the
module and the proven performance in cryogenic conditions (see tests discussed in Section |3|) led
us to choose this approach for our prototype.

Total weight is 533 grams, the rods and pins representing the 2.6% and 1.8%, respectively.

3. Testing

Electromagnetic tests were performed to validate the prototype and the manufacturing technique by
comparing measurements with electromagnetic simulations. In particular we measured the angular
response (beam patterns), insertion loss (IL) and return loss (RL), which are related to the scattering
parameters as follows:

\S I 2
7L = 10-log 10 i ! 2l J ; 7?L=-10-log 10 |5„| 2 . (3.1)

3.1 Beam Patterns

Beam patterns were measured at 100 GHz in the anechoic chamber of the "Istituto di Fisica del
Plasma" (IFP-CNR) in Milan with an ABmillimetre 8-350-4 VNA. They were obtained by con-
necting one horn at a time to one VNA port using a tapered rectangular-to-circular transition. The
signal, launched by the horn under test, was collected by a standard gain horn connected to the
second VNA port. The E and H orthogonal planes beam patterns were measured for all four horns,
while the 45 degrees plane and the relative cross-polarisation were measured for one horn.

The measured E-plane and H-plane patterns for all feed horns, compared with the simulations
(black line) described in JTo|], are shown in Figures |3] and |], respectively.

The four patterns are very similar, both in E and H planes, and in good agreement with sim-
ulations. At large angles (|0| > 65°) the horn response is below -40 dB and the signal to noise
ratio in the measurements degrades. The ~ 2dB deviation (at -20 dB level) between measurements

Figure 2. Top Left: top view of a single plate screwed inside the CNC. There are nine screws: four (black)
on the sides and five (gray) located in previously drilled rod-holes. After fixing the plate the four antenna
holes were milled. Top Right: corrugation details. Bottom Left: prototype assembling scheme. Bottom
Right: The completed prototype.

and simulations, around the first side lobes, could be due to mechanical manufacturing tolerances,
that are of the order of ~ 0.03 mm. A comparison of these results with those of an electroformed
equivalent horn could clarify the source of these small deviations.

In Figure |5] a comparison between the measured and simulated (black line) patterns of the 45
degrees plane of one horn is shown. The results are similar to the E and H planes cases.

Figure ^ shows the comparison between the measured (red line) and simulated (green line)
cross-polarisation patterns relative to the 45 degrees plane of one horn. Also in this case deviations
could be due to mechanical manufacturing tolerances.

3.2 Return Loss

The RL measurements were obtained at the Physics Department of the University of Milano-
Bicocca with an Agilent 8510c VNA, connecting one horn at a time to one VNA port by means of
a tapered rectangular-to-circular transition. The VNA was calibrated with a one-port calibration at
the interface between the horn and the transition. The measured return loss is below -30 dB on a
30% bandwidth, with a maximum matching of -42 dB at 100 GHz, resulting within the simulated
requirement (see Figure 0). Moreover the measurements shows high repeatability for all horns. We
recall that the simulation does not take into account resistive losses (see Sec. 3.3), which is com-

CO

CD

s

o
Q.

Angle (deg)

Figure 3. Comparison of measured and simulated E-plane beam patterns

CD
"D

10

-10
-20
-30

CD

S. -40

-50
-60

-70

-80

-

hornl H

■

•

norno n

■

-

■

-

■

if

!||

^k

i miixpii

| ■

■ IV

i

I

-100

-50

Angle (deg)

50

100

Figure 4. Comparison of measured and simulated H-plane beam patterns

patible with the deviations at the -40 dB level. Moreover the measurements show high repeatability
among all horns.

7-

CO

10

-10

-20

-30

CD

3
o
Q. -40

-50

-60

-70

I ' ' '

1 '

■

45 Co-polar Measured

r Simulated

■

•

■

■

■

•

^A

\f\v

VVl

' fr

\i\l\L '

■

■

-80
-100 -50 50 100

Angle (deg)
Figure 5. Comparison of the measured and simulated (black line) patterns of the 45° plane of one horn.

10

-10

-20

CD

r -30

CD

3
o
D. -40

-50

-60

-70

-80

o-polar
ross-Dolar

■

E

-H

■

— Simulation

■

■

l/r^sA \ \

u 1 ^—

■

. ,a/W

\ I // \

/ TV A

\a nj ft \

/ mVjj

VI

\I /llnll '

I ' Iff

W 1 '

■

-

■

-100

-50

Angle (deg)

50

100

Figure 6. The planes E, H, 45° co-polar and 45° cross-polar for one the horn. The green line represents the
simulation (from |P|) of the 45° cross-polar plane.

Another return loss measurement, performed after removing the central rod, yielded very sim-
ilar results, (see Figures § and|9|), indicating that it could be possible, in principle, to use this space

Figure 7. Return loss measurements compared with the simulation.

to build a module with an increased packing efficiency (horn aperture area/total area) of 0. 1 .

GHz

90 95

Figure 8. Comparison of the RL with (blue curve) and without (pink curve) the central rod, the curves are
very similar. This result allows, in principle, to substitute the central rod with another horn.

The RL was measured, for one horn, also in cryogenic conditions, after two cooling cycles at

~ -35

-50

75

80

GHz

90 95

Hornl 8T
Horn 3 8T
Horrr28T
Horn 4 8T

85

90

95

100

-50

105 110

Figure 9. Return loss measurements of the 4 horns without the central rod. The curves are very similar to
the curves obtained with all the nine rods and they show high repeatability for all the four horns.

about 4 K in a cryogenic chamber. The horn, inside the cryogenic chamber, was connected to the
VNA port by means of a waveguides chain (Figure [T^) composed by (from the horn to the VNA):
i) a tapered rectangular-to-circular transition, ii) a waveguide bend, iii) a straight steel waveguide
(acting as a thermal break), iv) a waveguide vacuum window and v) an external straight waveguide.

Figure 10. The cryogenic setup used for return loss measurements. Left: the waveguides chain inside the
cryogenic chamber connecting the prototype to the waveguide vacuum window. Right: the horn array with
thermometers, facing an ECCOSORB® AN 72 electromagnetic trap to absorb transmitted radiation.

The VNA was calibrated with a standard one-port calibration at the interface between the last
waveguide and the vacuum window. The prototype was positioned inside the cryogenic chamber
and connected to the waveguides chain. The RL was first measured at 300 K and then at 4 K. For

this reason the measurements, shown in blue and red in Figure 11, include the return loss of the

-10-

chain connecting the horn to the VNA. The waveguides chain RL is responsible for the oscillatory
signature. By applying a "gate" (i.e. a filter in the time domain) with the VNA, the effect of the
waveguide chain is removed from the measurements (green line in Fig. O).

90 GHZ 95

Figure 11. Comparison of the RL of one horn at 300 K (blue curve) and 4 K (red and green curves). The
blue and red curves include also the return loss of the (same) waveguides chain, that connects the horn inside
the cryogenic chamber to the external VNA. The waveguides chain RL is responsible for the oscillatory
signature. In the green curve, the waveguide chain was filtered with a gate in the VNA time domain. These
measurements do not indicate a major change in the return loss that would otherwise occur if there were
thermally induced changes in the array

It is worth noting that the filtered measurements are significantly better than those performed
in standard conditions, because the gate removes the effect of the signal reflected by the waveguide
chain. Considering the "gated" measurements, the return loss is still below -30 dB on a 30%
bandwidth, confirming the initial room temperature performance. In any case, both red and green
curves, do not indicate any mismatch due to possible thermally induced plate displacements. This
is an important result confirming the platelet technique reliability.

In Figure |1| one can notice a return loss increase in the 105- 110 GHz region between 300 K
and 4 K. This effect could be due to thermally induced changes in the waveguides chain, in the
absorbing trap and/or in the platelet array itself. It is reasonable to assume that the main contribu-
tion comes from the waveguides chain, that is composed by two golden copper waveguides and a
straight steel waveguide, acting as a thermal break. The steel waveguide, in particular, has a very
different thermal expansion coefficient compared to the other waveguide sections and to the platelet
array. Moreover the steel waveguide is the longest one and is subject to the largest thermal gradient
which causes the largest differential thermal contraction. These effects are likely to be the main
cause of the oscillatory signature. Also thermal effects in the absorbing trap and in the platelet

11-

array cannot be excluded but they are likely to be much smaller than those in the waveguides chain.

3.3 Insertion Loss

The Insertion Loss measurements were performed in transmission, with a TRL calibration, using
an Agilent 8510c VNA at the Physics Department of the University of Milano-Bicocca. The signal
was injected inside the horn throat and measured at its exit. The difference between the output and
injected signals yields the lost signal absorbed by the material because of its non-perfect electrical
conductivity and backscattered at the input (the return loss, that is negligible in this case). In
order to collect the output signal, we used a second twin horn, coupled by the aperture to the first.
Because two horns were used, a factor 2 must be taken into account in order to obtain the single
horn IL. Furthermore we simplified the mechanical interface by realizing a dedicated pair of "twin
horns" as a single two-port device constituted by two faced identical horns with the same design of

the array horns and manufactured according to the same technique (Figure 12).

Figure 12. The two halves, individually re-assembled, of the Twin Horn device, constituted by two identical
platelet horns with the same WCAM01 design, used to measure the insertion loss. In the IL measurements
the two horns were assembled together with four single rods of double length, with nuts present only on the
external sides.

The insertion loss measurements were obtained by connecting the two horns to the two Agilent
8510c VNA ports and measuring the 52,1 (or Sip) parameter with the same rectangular-to-circular
transitions used for the RL measurements. The VNA was calibrated with a TRL calibration at the
circular interface between the horns and the transitions.

The IL measurement is shown in Figure |13|, yielding an average value over the band of about -
0.06 dB per single horn. It is worth underlining that this figure represent an upper limit, because the
wave fronts of the two horns do not match perfectly, being spherical, an not planar, at their mouth.

12

This slight mismatch causes resonances between the horns that were removed by downsampling

the measurements (see Fig. 13) but yield, nevertheless, a slightly larger average insertion loss.

0,00
-0,05
-0,10
-0,15

CO

-o -0,20

CM

-0,25 -

-0,30

-0,35 -

-0,40

GHz

■Twin Horns Insertion Loss

110

0,0

-0,1

-0,2

-0,3

-0,4

80

90

100

110

Figure 13. Insertion Loss of the Twin Horns measured in transmission (S21). Data were resampled in order
to eliminate negative resonance spikes. The insertion loss of a single horn is the half of that shown in the
figure.

4. Conclusions

In this paper we have discussed the manufacturing and testing of a non-bonded W-band 2x2 array
prototype of corrugated feed-horns realised by "platelet" technique. Feed-horns have been realized
by layers, allowing reduced times and costs with respect to standard techniques like electroforma-
tion, typically adopted for frequencies > 100 GHz.

Beam patterns and RL measurements are in good agreement with the corresponding simula-
tions and show high repeatability for all the feed-horns, also in cryogenic conditions. In particular
the RL is less than -30 dB over a 30% bandwidth with a maximum matching of -42 dB at 100 GHz
for all four antennas. Also beam pattern measurements indicate very good repeatability and a low
level of side lobes and cross-polar polarisation (both lower than -25 dB). Moreover an estimation
of the IL indicates that it is not worse than -0.06 dB over the W band.

These results are very promising for the time and cost-effective mass production of large array

of high-performances corrugated feed-horns, as showed by [11]. The non-bonded solution, chosen

here and in [0] to assemble the module, allow simple manufacturing with the possibility to disas-
semble the module if needed and maintain electromagnetic performance at cryogenic conditions.
Further developments of this assembling solution are needed (and in progress) for increasing the
packing efficiency (horn aperture area/total area) to 0.3 and reducing the mass. These develop-

13-

ments will enable the realization of large, lightweight and high density focal planes, as required by
the next generation experiments for the precise measurements of the CMB polarisation.

5. Acknowledgments

The work described in this paper was carried out in the context of PRIN-MIUR 2006 and ASI
contract 1/016/07/0 "COFIS".

References

[1] J. C. Mather, D. J. Fixen, R. A. Shafer, C. Mosier, and D. T. Wilkinson, Calibrator design for the
COBE Far Infrared Absolute Spectrophotometer (FIRAS), The Astrophysical Journal, 512: 511-520,
1999 February 20.

[2] M. Bersanelli, D.Maino, A. Mennella, Anisotropies of the cosmic microwave background, 2002 La
Rivista del Nuovo Cimento, Vol. 25 Nr. 9.

[3] Wayne Hu, Martin White, A CMB polarisation primer, 1997 New Astronomy, 2 323.

[4] J.E. Carlstrom, J. Kovac, E.M. Leitch, C. Pryke, Status of CMB polarisation measurements from
DASI and other experiments, 2003 New Astronomy Reviews 47 953-966.

[5] L. Page et al., Three year Wilkinson Microwave Anisotropy Probe (WMAP) observations:
polarization analysis, 2007 Astrophys. J. Suppl., 170:335.

[6] C. Bischoff et al., First Season QUIET Observations: Measurements of CMB Polarization Power
Spectra at 43 GHz in the Multipole Range 25 < ell < 475, 2010, astroph-1012-3191.

[7] M. Kangas, K. Copsey, P. Lubin, A modular 100-GHz high-gain scalar corrugated nonbonded platelet
antenna, 2005 IEEE Antennas and Wireless Propagation Letters, vol. 4.

[8] R. W. Haas, Further Development of MMW and SMMW Platelet Feed Horn Array, 1994 ASP
Conference Series Vol. 75.

[9] F Villa, M. Bersanelli, N.Mandolesi, Design of Ka and Q band corrugated feed horns for CMB
Observations, Internal Report ITESRE, 188/1997.

[10] FVilla, V. Martorelli, WCAM prototype feed horn design, Internal Report INAF/IASF-BO, 463/2007.

[11] J. Gundersen, E. Wollack, Millimeter-wave Corrugated Platelet Feeds, Journal of Physics:
Conference Series 2009, 155 012005.

14